Near-infrared (NIR) spectroscopy has been developed as a noninvasive tool for the direct, real-time monitoring of glucose, lactic acid, acetic acid, and biomass in liquid cultures of microrganisms of the genera Lactobacillus and Staphylococcus. This was achieved employing a steam-sterilizable optical- ber probe immersed in the culture (In-line Interactance Systemt ). Second-derivative spectra obtained were subjected to partial least-squares (PLS) regression and the results were used to build predictive models for each analyte of interest. Multivariate regression was carried out on two different sets of spectra, namely whole broth minus the spectral subtraction of water, and raw spectra. A comparison of the two models showed that the rst cannot be properly applied to real-time monitoring, so this work suggests calibration based on non-difference spectra, demonstrating it to be suf ciently reliable to allow the selective determination of the analytes with satisfactory levels of prediction (standard error of prediction (SEP) , 10%). Direct interfacing of the NIR syst em to the bioreactor control system allowed the implementation of completely automated monitoring of different cultivation strategies (continuous, repeated batch). The validity of the in-line analyses carried out was found to depend crucially on maintaining constant hydrodynamic conditions of the stirred cultures because both gas ow and stirring speed variations were found to markedly in fluence the spectral signal.
Near-infrared spectroscopy: a tool for monitoring submerged fermentation processes using an immersion optical-fiber probe
TAMBURINI, Elena;
2003
Abstract
Near-infrared (NIR) spectroscopy has been developed as a noninvasive tool for the direct, real-time monitoring of glucose, lactic acid, acetic acid, and biomass in liquid cultures of microrganisms of the genera Lactobacillus and Staphylococcus. This was achieved employing a steam-sterilizable optical- ber probe immersed in the culture (In-line Interactance Systemt ). Second-derivative spectra obtained were subjected to partial least-squares (PLS) regression and the results were used to build predictive models for each analyte of interest. Multivariate regression was carried out on two different sets of spectra, namely whole broth minus the spectral subtraction of water, and raw spectra. A comparison of the two models showed that the rst cannot be properly applied to real-time monitoring, so this work suggests calibration based on non-difference spectra, demonstrating it to be suf ciently reliable to allow the selective determination of the analytes with satisfactory levels of prediction (standard error of prediction (SEP) , 10%). Direct interfacing of the NIR syst em to the bioreactor control system allowed the implementation of completely automated monitoring of different cultivation strategies (continuous, repeated batch). The validity of the in-line analyses carried out was found to depend crucially on maintaining constant hydrodynamic conditions of the stirred cultures because both gas ow and stirring speed variations were found to markedly in fluence the spectral signal.I documenti in SFERA sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
